[racket] Nesting evt handlers

The intent of `handle-evt' is that the handling procedure is called in
tail position with respect to `sync'. That tail guarantee is not
consistent with nesting.

We could tail that the handler is called in tail position only if it's
not nested, but we normally use `handle-evt' in cases where tail
behavior matters (e.g., a server loop), and we want to be notified when
it can't work; unexpected non-tail behavior is too subtle a bug. If
tail behavior doesn't matter, we use `warp-evt'.

Returning a handle evt from a guard evt is a different kind of wrapping
than putting a handle evt in a wrap evt. The unwrapping with a guard
evt happens before the `sync', while wrap-evt unwrapping happens after.

At Thu, 28 Mar 2013 16:28:38 +0100, Tobias Hammer wrote:
> Thanks, that's clear now.
> 
> Another question regarding events: Why is it forbidden to put a handle-evt  
> inside another handle- or wrap-evt?
> I couldn't come up with a situation where this could cause trouble.  
> Especially as it is allowed to nest handle-evts if i put a guard-evt or a  
> struct with prop:evt in between.
> 
> Tobias
> 
> 
> 
> On Thu, 28 Mar 2013 16:17:00 +0100, Matthew Flatt <mflatt at cs.utah.edu>  
> wrote:
> 
> > Sometimes, `sync' picks the first evt that it is given before even
> > trying the second evt enough to call the evt-generating function. That
> > is, `sync' looks far enough to see all arguments are `evt?', but
> > calling the evt-generator part of a guard evt is part of the attempt to
> > choose the specific guard evt, and that action doesn't happen if some
> > other evt succeeds first.
> >
> > I'll work on the docs to clarify.
> >
> > Yes, `proc' should be `generator'. But, `generator' also should be
> > changed, because "generator" now means something specific and
> > different.
> >
> > At Thu, 28 Mar 2013 16:08:09 +0100, Tobias Hammer wrote:
> >> Hi,
> >>
> >> i am trying to understand how nack-guard-evt works. Therefore i created
> >> the appended test program. Sometimes it stops with the error message and
> >> sometimes it prints the three values. It is clear to me what the error
> >> means and why it is raised.
> >> What i am not understanding is, why the nack generator is not run every
> >> time sync is called. I would have expected from the docs that it is run  
> >> on
> >> every sync.
> >> Can someone clarify why my assumptions are wrong and how nack-guard-evt  
> >> is
> >> expected to behave?
> >>
> >> Related question: The docs for the function talk abound 'proc', should  
> >> it
> >> be 'generator'?
> >>
> >> Tobias
> >>
> >>
> >>
> >> ;; -----------------------------
> >>
> >> #lang racket
> >>
> >> (define nack #f)
> >>
> >> (sync (handle-evt
> >>         always-evt
> >>         (lambda (_) 'ALWAYS))
> >>        (nack-guard-evt
> >>         (lambda (e)
> >>           (set! nack e)
> >>           never-evt)))
> >>
> >> nack
> >> (printf "~a\n" (sync/timeout 0. nack))
> >>
> >> > 'ALWAYS
> >> #f
> >> . . sync/timeout: contract violation
> >>    expected: evt?
> >>    given: #f
> >>    argument position: 2nd
> >>    other arguments...:
> >>     0.0
> >>
> >> OR
> >>
> >> > 'ALWAYS
> >> #<evt>
> >> #<void>
> >>
> >>
> >>
> >> --
> >> ---------------------------------------------------------
> >> Tobias Hammer
> >> DLR / Robotics and Mechatronics Center (RMC)
> >> Muenchner Str. 20, D-82234 Wessling
> >> Tel.: 08153/28-1487
> >> Mail: tobias.hammer at dlr.de
> >> ____________________
> >>   Racket Users list:
> >>   http://lists.racket-lang.org/users
> 
> 
> -- 
> ---------------------------------------------------------
> Tobias Hammer
> DLR / Robotics and Mechatronics Center (RMC)
> Muenchner Str. 20, D-82234 Wessling
> Tel.: 08153/28-1487
> Mail: tobias.hammer at dlr.de